Visual search has tripled in three years. It is brilliant at backpacks and shoes. It is much weaker at the thing many people point it at.
Typing used to be the only way to ask a search engine anything. That habit is fading faster than most people realise.
In May 2022, Google said its Lens tool was handling 8 billion visual searches a month, two and a half times the volume of a year earlier, AR Insider reported from that year's developer conference. By October 2024, Google's own blog put the figure at nearly 20 billion a month, and called Lens one of the fastest-growing query types on Search, with 18- to 24-year-olds using it most.
The latest number is higher again. Lens has passed 25 billion queries a month, with about one in five showing commercial intent, according to a Think with Google briefing cited by Scalevise.
Put that on a clock. Twenty-five billion a month works out to roughly nine to ten thousand visual searches every second, around the clock. In the time it takes to read this sentence, tens of thousands of photos have been turned into questions.
The growth curve matters as much as the total. From 8 billion to 25 billion is roughly a tripling in about three years, for a tool many people still think of as a party trick for identifying plants.
Lens no longer needs a camera, either. Google describes it as a way to search what you see "with your camera or on your screen", which means a screenshot of a feed, a frame from a video or an image in a group chat can all become a query in a couple of taps.
Visual search shines when the answer is a product. Google's example is a backpack spotted at an airport: take a photo, and the system matches it against a Shopping Graph of more than 45 billion products to find the exact item.
That works because products want to be found. Every listing is public, labelled and linked, so the photo has something to match against. Retailers spend heavily to make sure their images are indexed, tagged and current. The system is fed by the very people who want to be discovered.
People are a different problem. A photo of a person can lead to old social posts, to a model's portfolio, or to nowhere at all. It rarely leads to the page where that person actually publishes today, especially when that page sits behind a subscription.
The reason is structural. A search engine can only match what it can see, and paid content is, by design, not visible to it. The better a creator protects her work, the less a reverse image search can say about where it lives.
There is also a good reason it should stay that way. Identifying a stranger from a photo is exactly the use that privacy laws and platform rules are written to limit.
So for finding creators, the old method still works better than the new one. Names, categories and rankings carry information a picture cannot: who the person is, what she publishes and where she has chosen to be found.
Someone who has seen a face and wants to find where she posts is usually better served by a ranked list of the best models on OnlyFans, browsable by category and country, than by uploading a screenshot and hoping, with one catch: it covers only creators who have been added, and its order reflects popularity, not quality.
Visual search will keep growing. Twenty-five billion queries a month is a habit, not a novelty, and the youngest users are leading it.
But the growth is uneven. Pictures have become the fastest way to find a thing. Words remain the fastest way to find a person who chose to be found.
A camera can tell you what something is. It still cannot tell you who someone is, or where she wants to be seen.